Improving Septic Shock Prediction with AdaBoost and Cox Regression Model

Abstract

Septic shock in the advanced state of sepsis, which is a dangerous organ dysfunction disease that happens when the body responds in a dysregulated way to infectious diseases. Sepsis is hard to discover early on, and is difficult to treat if not detected sooner, hence, leading to high mortality rates. The efforts to improve the methods for identifying septic shock is ongoing in the medical and computer science communities. This paper uses the MMIC-III database to create a model to effectively predict septic shock utilizing a combination of the Cox regression model and AdaBoost. The prediction model is constructed by acquiring a risk factor score using Cox regression on various septic shock indicators. The score was appended as a feature to a selected listing of indicators and the AdaBoost ensemble classifier was applied to deliver the model. The predictive accuracy of the Cox Enhanced AdaBoost (CEAB) model was compared to prominent models to evaluate its effectiveness.